How Spite Can Save Fairness

Imagine two strangers on a desert island, arguing over a pile of coconuts. One of them owns the palm tree. The other provides the labor to crack them open. They must agree on how to split the harvest—or they both go hungry.

Now imagine this negotiation doesn't happen once, but dozens of times, in a world where the coconut trees slowly grow back after being stripped, but can also wither into stumps if harvested too aggressively. One round, both players might cooperate and share fairly. The next, one might lowball the other with an insultingly small offer. And sometimes—counterintuitively—spitefulness might actually be the behavior that saves the island.

This is the strange, counterintuitive world that physicists Arunava Patra, Prosanta Mandal, and Sagar Chakraborty from the Indian Institute of Technology Kanpur have spent years modeling. Their recent paper, published in 2026 on arXiv, doesn't just describe a hypothetical scenario. It identifies a precise mechanism by which spite—a behavior that hurts both the spiter and the person being slighted—can paradoxically become the engine of fairness. The key variable is the resource itself: how quickly it replenishes, how slowly it recovers. In a world of scarce, slow-growing resources, spite appears to be the unexpected pressure that forces the system back toward equity.

The finding challenges a long-standing assumption in evolutionary biology—that spiteful behavior, which carries a cost with no immediate reward, should be weeded out by natural selection. Instead, the researchers show that spite can survive, and even thrive, precisely because of what it does to the environment, not to the opponent.

The Science

To understand why this matters, you need to understand what evolutionary game theorists actually do. They treat social interactions—like bargaining, cooperation, or competition—as games, assign payoffs to different outcomes, and then ask: under repeated interactions and natural selection, which strategies win? The most famous example is the Prisoner's Dilemma, where two players each decide whether to cooperate or betray. The logic seems simple: betraying always yields a better individual payoff, so rational self-interest should win. And yet cooperation exists everywhere in nature.

The ultimatum game is another canonical framework, but one that probes fairness rather than cooperation. In its classic form, a proposer is given a sum of money and must offer a division to a responder, who can either accept or reject. If the responder rejects, both get nothing. Standard economic theory, built on rational self-interest, predicts the proposer will offer the smallest possible amount and the responder will accept it—because something is better than nothing. But that isn't what humans do. Across dozens of cultures and experiments, people consistently offer roughly 40-50% of the pot, and responders routinely reject lowball offers even when doing so earns them nothing. The implication is striking: humans are willing to punish unfairness even at a personal cost. They have a sense of fairness that defies simple optimization.

Patra and colleagues wanted to ask a deeper question. Most research on fairness and spite treats the game as happening in a static environment—a fixed pie, a fixed stakes size. But real life doesn't work that way. Resources grow, deplete, and recover. A forest regenerates after logging. A fishery collapses if overfished. A bacterial colony depletes its nutrients and then regrows. What happens when the game itself changes depending on what the players do to the world around them?

The researchers built a stochastic game—the stochastic mini-ultimatum game—where two players, an offerer and an accepter, repeatedly negotiate over a shared resource. The offerer owns the resource. The accepter brings something to the table—expertise, labor, the ability to actually extract value from it. In each round, the offerer proposes a division. The accepter can accept or reject. If they agree, the resource is harvested, depleting it slightly. If they reject, no harvest occurs, and the resource has a chance to recover.

The resource itself is modeled as self-renewing, governed by logistic growth—a standard way ecologists describe populations that grow quickly when abundant and slowly when depleted. To make the mathematics tractable, the researchers simplified the resource into two discrete states: a replete state, where the resource is abundant and healthy, and a depleted state, where it has been heavily used. Which state the resource is in changes the stakes of the game. In the replete state, a successful agreement yields the full normalized amount (set to 1 for simplicity). In the depleted state, successful agreements yield less—some fraction denoted by n, where 0 < n < 1. The environment is not a passive backdrop. It is a third player that responds to what the other two do.

The researchers then introduced strategy into this environment. Both the offerer and the accepter adopt reactive strategies—so named because each player's action depends on what the other player did in the previous round. This is a cognitively simple rule, requiring only memory of the last interaction, and it closely mirrors how real repeated negotiations unfold: you adjust your offer based on whether your partner accepted or rejected last time.

Four pure reactive strategies are possible in each resource state. The fair strategy, denoted F, involves always choosing the equitable option regardless of history. The unfair strategy, U, involves always proposing or demanding the lopsided option. The complier strategy, C, involves matching the other player's previous action—complying with whatever they did last time. And the anti-complier strategy, A, involves doing the opposite—anti-complying with the other player's last move.

Because the resource can be in either state, each player effectively has a pair of strategies: one for the replete state, one for the depleted state. The space of possible strategy combinations is therefore large, but still tractable enough to analyze systematically.

The researchers then studied the evolutionary dynamics using a mutation-selection framework. Imagine a large population of offerers and a large population of accepters, each playing the game repeatedly. Periodically, a player might randomly mutate to a different strategy. Natural selection then favors strategies that yield higher average payoffs over time. The key parameter here is the discount factor, δ—the weight placed on future payoffs relative to present ones. A high δ means players care a lot about maintaining a good relationship and future opportunities. A low δ means they prioritize short-term gains and are more willing to exploit or punish.

The researchers simulated populations of varying sizes—between 10 and 100 individuals per group—and computed the equilibrium frequencies of each strategy combination, tracking how often fairness, spite, and the resource's state appeared across different conditions.

What They Found

The results fall into two broad regimes, and the dividing line is the resource growth rate—a parameter that captures how quickly the environment replenishes itself after depletion.

In the first regime, corresponding to fast-growing resources, fairness is dominant. When the environment bounces back quickly (modeled by the transition vector 𝝉₁₁₁₁), the resource spends most of its time in the replete state, and fair strategies—F for the offerer and F for the accepter—prevail. In this world, there is enough to go around. The offerer doesn't need to cheat, because the resource is abundant enough that sharing fairly doesn't feel costly. The accepter doesn't need to punish, because the next round will offer a fresh opportunity anyway. The researchers' simulations, varying the discount factor δ between 0.1 and 0.99 and population sizes between 10 and 100, consistently showed high fairness levels in this regime. Fairness levels climbed toward 1 as the discount factor increased, especially in larger populations.

But this is the expected result. It confirms something we might already have intuited: abundance promotes generosity. What the paper reveals—its genuinely surprising contribution—is what happens in the second regime: with slow-growing resources.

When the resource recovers slowly (modeled by the transition vector 𝝉₀₀₁₀), the system behaves very differently. Here, the depleted state becomes a trap. Successful agreements—fair or unfair—deplete the resource further and keep it stuck in the depleted state. The resource can only escape depletion through a specific kind of failure: the spiteful outcome, where a low offer (L) meets a high demand (H). This is the (L, H) action pair. In the mini-ultimatum game framework, (L, H) is spiteful precisely because neither player gets anything—the offerer lowballed, the accepter refused, and both walk away empty-handed. No harvest means no depletion. And crucially, in this slow-growth regime, the resource then has time to recover.

This creates a feedback loop. When the resource is depleted, the best thing both players can do—for purely selfish reasons—is act spitefully. The spiter gets no payoff this round, but by refusing to agree, they force the resource to recover enough for a better deal next time. This self-interested spitefulness drives the resource back toward the replete state. And once the resource is replete, something remarkable happens: the repeated interaction mechanism kicks in, and fairness becomes evolutionarily stable. The offerer who has been behaving spitefully in the depleted state now faces an accepter who has learned to expect better. And in the replete state, the incentive shifts: fair agreements yield real payoffs, and the shadow of future interactions encourages both players to honor that shift.

The researchers documented this feedback loop quantitatively. For the slow-growth resource, spite levels were elevated in the depleted state, and resource replete levels showed a non-monotonic relationship with the discount factor—a distinctive signature of the feedback mechanism. As the discount factor increased, the system showed complex patterns: moderate δ values produced mixed equilibria with both spite and fairness present, while very high δ values drove convergence toward fairness, but through a route that first required spite to rescue the resource.

The histogram in Figure 4 of the paper provides perhaps the most striking illustration. It shows the distribution of outcome profiles—the combination of what happens in the replete and depleted states—across many simulation runs. For the slow-growth resource (transition vector 𝝉₀₀₁₀), the distribution was broad and heterogeneous. Different populations settled into different attractors, with both fair and spiteful equilibria observed simultaneously. This is not a tidy, predictable world. It is one where history matters, where the initial conditions of a negotiation determine whether a population ends up in a fair equilibrium or a spiteful one. And crucially, it is one where spite is not a bug but a feature of the system's self-correcting mechanism.

To make this concrete: in the slow-growth regime, the researchers found that the anti-complier strategy for the offerer (Aₒ)—essentially, making the opposite move to whatever the accepter demanded last—paired with the fair strategy for the accepter (Fₐ), produced a robust equilibrium. The offerer's anti-complier behavior generated low offers in the depleted state, which triggered spiteful rejections, which allowed the resource to recover, which eventually shifted the game to the replete state where fairness could take hold. Spite, in other words, was doing ecological work that fairness alone could not do.

Why This Changes Things

Evolutionary biology has long struggled with spite. Unlike cooperation, which can be explained by direct benefits, kin selection, or reciprocity, spite poses a puzzle: why would you harm yourself to harm someone else? The classic answer is that spite can serve as a signal—perhaps of commitment, or of willingness to reject bad deals—or that it can function through indirect reciprocity, where harming a low-reputation individual improves your standing with onlookers.

The IIT Kanpur team's contribution is to place spite in an ecological context. Spite is not just a social signal. It is an environmental management strategy. When resources are scarce and slow to recover, spiteful behavior is the mechanism by which the system forces restraint. By refusing to participate in exploitative deals, both parties buy time for the resource to recover. The spiteful rejection, which looks irrational from a short-term payoff perspective, is actually a form of collective patience—one enacted through mutual punishment rather than mutual trust.

This reframing has implications that reach well beyond game theory. Consider fisheries. A fishing community that always harvests at maximum capacity may individually maximize their short-term catch, but collectively they drive the stock to collapse. A norm of refusing bad deals—in which boat owners demand fair shares of the catch and workers refuse exploitative wages—might seem like spiteful behavior that hurts everyone in the short run. But if it reduces the pressure on the stock, it creates the conditions for long-term sustainability. The fairness that eventually emerges is built on the back of what looked, initially, like mutual spite.

Or consider labor markets. A worker who refuses a wage offer that undervalues her labor might be acting spitefully—she incurs a short-term cost to deny the employer the benefit of cheap work. If many workers do this, employers are forced to raise wages, creating more equitable outcomes. The mechanism is not cooperation. It is a coordinated spite response that reshapes the economic environment, which then makes fairness viable.

The paper also speaks to a broader debate about rationality. Classical economics assumes that rational self-interest guides decision-making, which would predict that in the ultimatum game, proposers should offer the minimum and responders should accept any positive amount. But neither humans nor, as this model shows, evolutionarily stable strategies follow this prediction. Fairness exists. Spite exists. And the IIT Kanpur model explains how both can be stable outcomes of the same underlying dynamic—not as contradictions, but as complementary phases of a system that is collectively managing a shared resource.

Another important dimension is the role of ownership. The model assigns the resource to the offerer, who must share it to get anything. This setup prevents the tragedy of the commons—the well-known problem where shared resources get overexploited because no individual bears the full cost of depletion. Ownership creates accountability. But ownership alone isn't sufficient. What the paper shows is that ownership, combined with repeated interactions and the right resource dynamics, creates a framework where fairness can evolve endogenously—not because players are inherently fair, but because the structure of the game rewards fairness in the long run.

The finding that resource growth rate determines which regime a population inhabits is also notable. It implies that environmental factors aren't just constraints on social behavior—they are active shapers of moral dynamics. A population with fast-recovering resources may develop a culture of fairness almost effortlessly. A population with slow-recovering resources may need to pass through a phase of apparent spitefulness before fairness can take hold. This is not a comforting story about human nature being inherently cooperative. It is a more nuanced story: the same species, facing different ecological conditions, may evolve different moral economies.

What's Next

The paper opens several threads that future research could pull. The model uses a highly simplified two-state resource and a cognitively minimal reactive strategy. Real resources are continuous, and real negotiations involve richer memory and more sophisticated reasoning. Extending this framework to continuous resource dynamics and longer memory horizons would test whether the feedback mechanism the authors identified is robust to more realistic assumptions—or whether it dissolves in the complexity of the real world.

The researchers themselves note that the stochastic game framework they introduce could be adapted to study how learning and memory shape the evolution of spite and fairness in a fluctuating environment. A population that can learn from recent history might develop adaptive strategies that smooth out the oscillation between spite and fairness. There is a potential for machine learning approaches here: how does a reinforcement learning agent behave in this stochastic ultimatum game? Does it discover the spite-driven feedback loop, or does it converge more quickly to fairness through better modeling of the resource dynamics?

Another open question concerns the transition between regimes. The paper identifies two distinct regimes based on resource growth rate, but natural resources rarely have fixed growth rates. Climate change, habitat loss, and human intervention can alter growth rates in real systems. What happens at the boundary between regimes? Does the system transition smoothly, or does it undergo a phase transition—abruptly flipping from a fairness-dominant to a spite-dominant equilibrium as the resource slows down? Understanding this transition could have practical implications for managing ecosystems on the brink of collapse.

A further limitation is the assumption of a single owner and a single laborer. Most real resource negotiations involve multiple owners, competing bidders, or communal ownership structures. How the feedback mechanism operates in multiplayer settings, or under different ownership norms, remains unexplored. Similarly, the paper does not address spatial structure—real populations are not well-mixed; individuals interact in networks, with neighbors having more frequent contact than strangers. The authors note that network reciprocity has been studied in other contexts and could be incorporated here.

There is also the question of empirical validation. The model makes predictions about the relationship between resource dynamics and the prevalence of spite and fairness. Testing these predictions would require long-term data from real populations facing different resource regimes—perhaps comparing fishing communities that harvest slow-growing species versus fast-growing ones, or agricultural communities that manage fast-recovering versus degraded land. The theoretical machinery exists. The empirical bridge is yet to be built.

Perhaps the most profound open question is whether the spite-driven feedback mechanism generalizes beyond resource management. Could the same logic apply to information ecosystems, where spiteful rejection of bad-faith arguments creates space for better discourse? Could it apply to political systems, where a mobilized minority's refusal to cooperate with unjust arrangements forces structural reform? The abstract structure of the model—a shared, depletable resource, contested through repeated negotiation—maps onto many domains beyond biology. The authors have built a theoretical scaffold. What hangs from it, in the world outside the model, is still being discovered.

What is clear is that Patra, Mandal, and Chakraborty have demonstrated something important: that the evolutionary origins of fairness and the puzzle of spiteful behavior are not separate problems to be solved independently. They are two aspects of the same story—one written not in the language of morality or rationality, but in the language of ecology. Spite and fairness are not opposites fighting for dominance. They are partners in a negotiation between a population and its world.